Published ahead of print on April 7, 2004, doi:10.1164/rccm.200309-1239OC
© 2004 American Thoracic Society A Simplified Method for Measuring Critical Pressures during Sleep in the Clinical SettingDepartment of Medicine, Division of Pulmonary and Critical Care Medicine, Johns Hopkins University, Baltimore, Maryland Correspondence and requests for reprints should be addressed to Susheel P. Patil, M.D., Johns Hopkins Sleep Disorders Center, Asthma and Allergy Building, 5501 Hopkins Bayview Circle, Room 4B.50, Baltimore, MD 21224. E-mail: spatil{at}jhmi.edu
Upper airway critical pressure measurements correlate with the degree of upper airway obstruction during sleep and may have a role in the diagnosis and treatment of obstructive sleep apnea. Nevertheless, the utility of the critical pressure has not yet been realized in the clinical setting because significant technical expertise is still required for the acquisition and analysis of pressure–flow data. Using segmented regression, we developed and validated a simplified approach to analyze the pressure–flow relationship and to determine the effects of protocol-related factors in 44 subjects with sleep apnea. When compared with expert visual analysis, segmented regression method was found to accurately determine the critical pressure (–0.98 ± 2.47 cm H2O vs. –1.07 ± 2.47 cm H2O, respectively; p = 0.46). Furthermore, it was found that two series of measurements acquired at varying nasal pressure levels with two or more breaths per level were sufficient to determine the critical pressure with a minimum of variability. Therefore, this analytic approach has the potential for standardizing and simplifying the ascertainment of the critical pressure for studies examining the effect of therapeutic devices and agents on upper airway collapsibility during sleep.
Key Words: sleep apnea • critical pressure • pathophysiology • upper airway Obstructive sleep apnea is a common disorder that is characterized by repetitive collapse of the upper airway during sleep. Obstruction of the upper airway has been attributed to increased pharyngeal collapsibility that may be related to alterations in either structural and/or neuromuscular properties of the upper airway (1, 2). Measurements of upper airway collapsibility (i.e., critical pressure) have been shown to correspond with the degree of airflow obstruction in individuals who have complete, partial, or no airflow obstruction during sleep (3). These measurements may help elucidate the pathophysiology of obstructive sleep apnea, identify individuals who are at risk for sleep apnea, and guide therapy for the disorder (4–9). Despite the potential, measurements of upper airway collapsibility have not yet been incorporated into clinical practice, in part because significant technical expertise is required to implement protocols for determining critical pressure during sleep. The critical pressure is determined by altering nasal pressure systematically during sleep (1, 2, 10–14) and is defined by the nasal pressure below which the upper airway occludes and airflow ceases. Nevertheless, studies to date differ substantially in the protocol used to alter nasal pressure and in the methods for analyzing pressure–flow data generated (discussed later here). Thus, a lack of standards for data collection and analysis has impeded the development of uniform methods for determining critical pressure and may introduce variability in results between study populations and sleep centers. We have previously developed an abbreviated protocol for delineating the upper airway pressure–flow relationship in a small sample of subjects with obstructive sleep apnea (15). The protocol standardized the exposure to nasal pressure but still required substantial expertise in the collection and analysis of pressure–flow signals for accurate determinations of critical pressure. Specifically, esophageal manometry and pressure–flow waveform analysis were required to identify the subset of "flow-limited" breaths to be included in the analysis. Thus, despite standardization of the data acquisition protocol, significant expertise was still required to perform esophageal manometry and to analyze pressure–flow data. The purpose of this study was to develop and validate a simplified, noninvasive method to identify the flow-limited segment of the pressure–flow relationship during sleep. The flow-limited segment is characterized by a positive slope of the pressure–flow relationship, as distinguished from segments with either an indeterminate or zero slope (non–flow-limited or occluded segments) (1, 12). Analytic methods were developed to identify the sloped portion of the pressure–flow relationship over which the airway collapses and flow limits. It was hypothesized that the flow-limited (sloped) portion of the pressure–flow relationship can be accurately identified with the method of segmented regression (16–22), which is well suited for modeling changes in slope that correspond to distinct states of upper airway patency (occluded, flow-limited, and non–flow-limited). It was further hypothesized that simplifying and standardizing methods for data acquisition and analysis would allow us to account for protocol-related factors, such as the number and duration of nasal pressure levels that could increase variability in the determination of critical pressure.
Theoretic Approach. The Starling resistor model has been previously used to describe upper airway pressure–flow relationships (1, 12). As predicted by the model, the upper airway flow-limits when downstream pressure falls below the critical pressure. Under flow-limited conditions, pressure downstream to the site of collapse no longer influences maximal inspiratory airflow (  Imax). Rather, Imax varies solely with changes in the upstream nasal pressure (23) and increases proportionately with elevations in nasal pressure, leading to a sloped pressure–flow relationship. The lower end of the flow-limited (sloped) segment is bound by critical pressure, the nasal pressure below which airflow ceases (Figure 1
, segment A). The upper end of the pressure–flow relationship is bound by the minimally effective therapeutic pressure (Peff), the nasal pressure above which airflow limitation is abolished (Figure 1, segment C). As nasal pressure continues to rise above the minimally Peff, the downstream pressure at peak inspiration no longer falls below critical pressure, and a flow-limited condition no longer obtains. Under these circumstances, Imax is determined by the gradient, nasal pressure – downstream pressure, which remains relatively constant during stable nonrapid eye movement sleep over a wide range of nasal pressure (24, 25).
The flow-limited segment of the pressure–flow relationship is used to define the critical pressure as the nasal pressure at which airflow ceases. The critical pressure is derived from linear regression of the data comprising the flow-limited segment (Figure 1, segment B). Using previously described methods for data acquisition (15), we assessed Imax through the upper airway during sleep over a range of nasal pressure. Discrete levels of nasal pressure were set (bins), and maximal airflows were measured for several breaths at each nasal pressure level.
Analytic Approach.
Modeling the flow-limited segment.
(
Patient Recruitment Forty-four subjects with obstructive sleep apnea (apnea–hypopnea index  20 events/hour) presenting for continuous positive airway pressure titration were studied. Sleep apnea severity was determined by overnight polysomnography as previously described (27). Subjects were divided consecutively between a development sample (sample A; n = 30) and a validation sample (sample B; n = 14). The study was approved by the institutional review board on human research.
Experimental Protocol During sleep, nasal pressure was maintained at a holding pressure that eliminated flow limitation (15). Nasal pressure was abruptly lowered for five breaths (a run) through a remote-control device attached to a continuous positive airway pressure unit designed to apply pressures between –20 cm to 20 cm H2O. Three series of stepwise reductions in nasal pressure that encompassed zero airflow (critical pressure) were collected (Figure 2) . If an arousal occurred, the protocol was resumed after patients reinitiated stages II–IV nonrapid eye movement (NREM) sleep. Breaths associated with microarousals from sleep were excluded from analyses.
A recording example of pressure–flow measurements is shown in Figure 3A for one series of runs at several nasal pressure levels during stable NREM sleep, with corresponding median Imax versus nasal pressure plot (Figure 3B). Imax was measured as the difference in inspiratory flow maximum and the zero or mean airflow level (average airflow between the peak expiratory flow of breaths) during a run. A Imax versus nasal pressure plot was generated for each subject.
Sensitivity Analyses Results of the segmented regression analyses for each individual were compared against the visual identification approach of identifying the flow-limited segment. Two experts (H.S. and A.R.S.) independently examined each pressure–flow curve, and identified the flow-limited segment. Separate analyses were undertaken to compare values of critical pressure and RUS obtained from an analysis of (1) only flow-limited breaths, based on established criteria (28, 29), and from an analysis of (2) all flow-limited and non–flow-limited breaths. If significant differences in critical pressure or RUS were not detected between these analytic methods, we concluded that flow-limited breaths need not have been identified visually and that the detection and analysis of the flow-limited segment could be automated. Sensitivity analyses were performed to optimize thresholds for the minimal slope and no-flow criteria by testing specific combinations of different thresholds on sample A only. A comparison of segmented regression to expert visual identification demonstrated that a minimal slope criterion of 20 ml/second/cm H2O and a no-flow threshold of 50 ml/second had the greatest levels of agreement (see the online supplement, section B). The selected criteria were used to determine critical pressure and RUS in both samples using segmented regression and prospectively validated in Sample B only.
Effects of Protocol-related Factors on the Upper Airway Pressure–Flow Relationship
Statistical Analyses
Forty-four subjects with obstructive sleep apnea were studied and divided between a development (sample A) and validation sample (sample B). The first 30 patients (sample A) were used to determine the minimal slope criteria and no-flow threshold criteria, whereas the final 14 patients (sample B) were used to validate the selected criteria. Patient characteristics of the development and validation samples were comparable in age (48.3 ± 9.8 vs. 47.4 ± 8.9 years), body mass index (36.6 ± 6.7 vs. 35.9 ± 9.1 kg/m2), NREM apnea–hypopnea index (68.3 ± 27.0 vs. 75.1 ± 31.6 events/hour), and sex (73.3 vs. 85.7% male). There were no significant differences in characteristics between the two groups. The median duration of time required to acquire pressure–flow data on each individual was 43.5 minutes (interquartile range of 30.0–68.8 minutes). The median number of runs required to obtain the three series of pressure–flow data was 17 runs (interquartile range of 14–23 runs). The median number of arousals during acquisition of data was 3 (interquartile range of 1–7). For each patient, individual pressure–flow relationships were constructed. In the segmented regression analyses of patients in sample A, a minimal slope criterion of 20 ml/second/cm H2O and a no-flow criterion of 50 ml/seconds demonstrated the greatest agreement in critical pressure and Peff with the expert visual analysis (see Tables E1 and E2 in the online supplement). Using the identified criteria, we compared the expert visual analysis to the segmented regression analyses and found no significant difference in critical pressure (–0.98 ± 2.47 vs. –1.07 ± 2.47 cm H2O, p = 0.46), Peff (6.60 ± 3.49 vs. 6.92 ± 3.37 cm H2O, p = 0.40), and RUS (17.28 ± 9.65 vs. 17.25 ± 8.28 cm H2O/ml/second, p = 0.97), respectively, between the two methods within Sample A. Bland-Altman plot is illustrated for sample A in Figure 4A . The plot demonstrates that the critical pressure derived by the segmented regression analyses did not differ systematically from that derived by the expert visual analysis (mean difference of 0.09 cm H2O; 95% confidence intervals [CI], –0.15 to 0.32 cm H2O) and that the limits of agreement were narrow (lower limit of agreement –1.18 cm H2O; 95% CI, –1.59 to –0.77 cm H2O; and upper limit of agreement 1.35 cm H2O; 95% CI, 0.94 to 1.76 cm H2O). Outliers generated from the segmented regression were minimal.
The selected minimal slope and no-flow criteria were then validated in sample B. A comparison of the expert visual with the segmented regression analyses revealed no significant difference in critical pressure (–0.70 ± 3.07 vs. –0.76 ± 3.12 cm H2O, p = 0.78), Peff (7.56 ± 4.65 vs. 7.91 ± 4.21 cm H2O, p = 0.17), and RUS (16.32 ± 10.07 vs. 16.73 ± 9.90 cm H2O/ml/second, p = 0.52), respectively. A comparison of critical pressure determinations in sample B between the segmented regression method and expert visual analysis also demonstrated no significant systematic difference exists between the two methods (mean difference of 0.06 cm H2O; 95% CI, –0.40 to 0.52 cm H2O) and that the limits of agreement were narrow (lower limit of agreement –1.54 cm H2O; 95% CI, –2.35 to –0.74 cm H2O; and upper limit of agreement 1.66 cm H2O; 95% CI, 0.86 to 2.47 cm H2O), with only one outlier detected in the Bland-Altman analysis (Figure 4B). Analyses were also performed in both samples to compare differences in critical pressure and RUS obtained when using all breaths (flow-limited and non–flow-limited) against flow-limited breaths only. No significant differences in critical pressure (difference of –0.2 cm H2O, p = 0.12) or RUS (difference of –0.03, p = 0.97) were found in these analyses.
The influence of protocol-related factors on the upper airway pressure–flow relationship for the entire group was then examined. In Table 1
, interbreath differences in airflow across the five breaths after abruptly lowering nasal pressure are presented. For example, the change in
Subsequently, the minimum number of series of pressure–flow data necessary to determine critical pressure accurately was determined in the subset of subjects with at least three series of measurements (n = 34). Bland-Altman plots did not reveal any significant systematic differences between the two measurement methods for one series, two series, or three series of pressure–flow measurements. Qualitative inspection of Bland-Altman analyses, however, demonstrated a narrower reference range (±2 SD) for between-paired measurements (3.9 vs. 5.8 cm H2O) for two series compared with one series of pressure-flow measurements, respectively. In addition, a lower mean difference in between-paired measurements (–0.03 vs. –0.3 cm H2O) was determined for two series compared with one series of pressure–flow measurements, respectively. No additional improvement in the determination of critical pressure with three series was evident based on visual inspection (see Figures 5A–5C) , suggesting that a minimum of two series of pressure–flow measurements, which include inspiratory flows below 50 ml/second, were necessary to accurately determine critical pressure.
The objective of this study was to develop a systematic approach for the acquisition and analysis of pressure–flow data used for determining critical closing pressures during sleep. A simplified noninvasive approach to the assessment of upper airway pressure–flow relationship was presented and provides a valid analytic method to identify the subset of measurements that describe the functional properties of the upper airway. A major finding of this study was that segmented regression techniques could accurately identify the flow-limited segment and that critical pressure values based on automated analysis did not differ from those determined by expert visual observations. The results of this study also illustrate that a quasi steady state in the pressure–flow relationship was established within two breaths after changes in nasal pressure. Finally, the analyses demonstrate that at least two series of pressure–flow data were required to assess critical pressure accurately. Collectively, the results of this study provide a systematic approach for the acquisition and analysis of pressure–flow data used in determining critical pressure during sleep. Our approach to analyzing the pressure–flow relationship was chosen based on fundamental concepts regarding the pathophysiology of airflow obstruction in obstructive sleep apnea. As the nasal pressure is raised progressively in a patient with an occluded upper airway, three distinct states of upper airway patency have been observed (1, 12, 23, 25). Airflow ceases (the airway occludes) when nasal pressure is less than critical pressure, increases linearly during flow-limited breathing as pressure is raised above critical pressure, and becomes indeterminate with further increases in nasal pressure (above Peff) once flow limitation is abolished. Although a continuous sigmoid function could describe the pressure–flow relationship, it would not have been appropriate because discrete states of upper airway occlusion, flow limitation, and non–flow-limited breathing would not be modeled, as predicted by the Starling resistor model (31–33).
A further advantage of our analytic approach was that it allowed us to probe for methodologic and physiologic sources of variability in critical pressure determinations. Statistical methods were used to model repeated measures of breaths at each nasal pressure level, along with summary statistics (e.g., median flow) to characterize the pressure–flow relationship. We found that two complete series of pressure–flow data were sufficient to determine critical pressure accurately, and that a third series did not further increase our accuracy. In contrast, investigators have previously chosen the number of runs and series to be acquired (13–15, 34), which may have differed substantially within and between studies. Such differences may have influenced the accuracy and precision of critical pressure and may lead to bias in the estimates of critical pressure. Our methods also allowed us to account for breath effects, which may be secondary to changes in lung volumes (35–37). Others and we have previously shown critical pressure to decrease over several breaths during each run (14, 38, 39). In the repeated-measures analyses examining the effect of successive breaths on critical pressure, no statistically significant differences in A major focus of this study was to develop an approach that minimized the technical expertise required for the collection and analysis of pressure–flow data without compromising our ability to determine the critical pressure accurately. In previous studies, laborious methods were employed to ensure that sufficient flow-limited breaths were available to delineate the entire pressure–flow relationship. Esophageal manometry and visual inspection of pressure-flow waveforms (1, 12, 15, 23, 39) were required to extract the subset of flow-limited data to be used in determining critical pressure. Investigators frequently had to acquire additional pressure–flow data to ensure that sufficient flow-limited breaths would be ultimately available for analysis. In contrast, we implemented a uniform collection protocol that did not require invasive monitoring of esophageal pressure, which can disrupt sleep. We found that a complete data set could be obtained with only two series of runs over less than 1 hour of sleep data acquisition and that critical pressure could be accurately identified with segmented regression analysis without preselecting flow-limited breaths for analysis. Thus, our analytic approach eliminated the need for invasive monitoring of esophageal pressure, further simplified the data acquisition protocol, and reduced the technical expertise required for identifying the flow-limited segment and determining the critical pressure. Several pitfalls should be acknowledged that might limit the implementation of this approach in the clinical setting. First, the acquisition protocol required complete polysomnography and quantitative measurements of airflow and nasal pressure to generate pressure–flow data for analysis. Nevertheless, equipment for monitoring pressure and airflow is currently supported by many continuous positive airway pressure devices and polysomnographic recording systems. Second, personnel trained to detect stable sleep and microarousals in real time during data acquisition were required. Third, esophageal manometry was not used to identify flow-limited breaths, which might have allowed for more accurate identification of flow-limited breaths. Nevertheless, reliable methods for determining the presence of flow limitation based on visual inspection of flow waveforms have been established (28, 29). No significant differences in critical pressure were found when the technique of segmented regression was applied to datasets that included all flow-limited and non–flow-limited breaths or only flow-limited breaths. Further development of automated procedures for sleep stage monitoring and the detection of flow limitation on a breath-by-breath basis may overcome these limitations and more fully automate methods for determining the critical pressure in the clinical setting. A potential limitation of the presented methods is that pressure–flow measurements were generated under hypotonic conditions (14). Critical pressure measured with this protocol is primarily thought to reflect the influence of anatomic factors on upper airway collapsibility. Previously, investigators have demonstrated that critical pressure in normal individuals may be somewhat higher (less negative) under hypotonic conditions (13) than those determined under state conditions of intact neuromuscular activity (12). In contrast, the critical pressure under hypotonic conditions in our population of sleep apnea subjects appears to be remarkably similar to that described in the atonic state by Isono and colleagues (2). Thus, our protocol may be useful in examining the anatomic correlates of upper airway collapsibility that are seen in different populations (e.g., men vs. women) (40); however, our analyses will need to be extended to steady-state pressure–flow relationships assessed when neuromuscular activity is intact. Our findings have several implications for investigators seeking to characterize upper airway pressure–flow relationships and to determine critical closing pressures during sleep. First, an abbreviated protocol has been established for generating upper airway pressure–flow relationships and consists of two series of stepwise reductions in nasal pressure for at least two breaths at each pressure level. Because less than 1 hour was required to obtain two series of runs, our findings indicate that upper airway function can be accurately characterized in the clinical setting during a routine continuous positive airway pressure titration study. Second, technical expertise in the recognition of flow-limited breaths is no longer required to determine the critical pressure, thereby simplifying the analysis of acquired data to determine the critical pressure and the minimally Peff. Third, standardizing the data acquisition and analytic methods for determining critical pressure lays the foundation for future studies examining measurements of upper airway collapsibility across clinical populations and centers. In particular, the analytic methods have potential applications for clinical trials examining the effect of therapeutic devices and pharmacologic agents on upper airway collapsibility during sleep. Finally, the minimally Peff could be automatically determined from the pressure–flow relationship and might be used during nasal continuous positive airway pressure titration in sleep apnea patients. Further work is required to compare results obtained with the presented approach across study centers and cohorts.
The authors thank Mr. Luis Pichard and Ms. Elizabeth Gladmon for their assistance in the preparation of graphs and tables.
Supported by HL68481, HL50381, HL37379, HL04065, HL75078, M01-RR-02719 (General Clinical Research Center), and Medizin-Technologie fuer Arzt und Patient/Resmed, Inc. This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Conflict of Interest Statement: S.P.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; N.M.P does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; H.S. declares that the technology used in this study is under commercial development by Medizin-Technologie fuer Arzt und Patient, GmbH (MAP), and is a paid consultant to MAP, and the terms of this arrangement are being managed by the John Hopkins University in accordance with its conflict of interest policies; C.P.O. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; P.L.S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.R.S. declares that the technology used in this study is under commercial development by Medizin-Technologie fuer Arzt und Patient, GmbH (MAP), and is a paid consultant to MAP, and the terms of this arrangement are being managed by the John Hopkins University in accordance with its conflict of interest policies. Received in original form September 7, 2003; accepted in final form April 5, 2004
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
Conceptual Background
j Equation 1 In this equation, 
